NVIDIA Agent Toolkit is framework-agnostic and supports all major agentic frameworks without code changes. Full native support includes LangChain, LlamaIndex, CrewAI, Microsoft Semantic Kernel, and Google ADK. The toolkit also works seamlessly with AutoGen, AWS Strands, and custom Python agents. This flexibility means you can use your preferred framework—LangGraph for stateful orchestration, Deep Agents for complex reasoning, or simple prompt-loop agents—and layer the toolkit’s observability and optimization on top.
Framework-specific plugins are available as optional extras. For example, nvidia-nat[langchain] installs all LangChain and LangGraph integrations needed for advanced tracing, evaluation, and optimization. The Agent Performance Primitives (APP) accelerate graph-based frameworks with parallel execution, speculative branching, and node-level priority routing. Each framework integration maintains full backward compatibility, so existing code runs unchanged while unlocking profiling, cost optimization, and multi-agent coordination through A2A Protocol.
For semantic retrieval, all frameworks can integrate Milvus as a vector store for RAG. Milvus documentation shows integration examples across LangChain, LlamaIndex, and other popular frameworks, enabling agents to retrieve context from enterprise knowledge bases.